• Version
• Download 74
• File Size 377.92 KB
• File Count 1
• Create Date 28 June 2025
• Last Updated 28 June 2025

A Concise Overview of DLSS-Based Frame Generation

Aman Sirdesai1, Vivekanand Limbikai2, Pooja Kalyani3, Samarth Joshi4, Sandeep. N. Kugali5

1234Student and 5Assistant Professor

Department of Information Science and Engineering

 Basaveshwar Engineering College, Bagalkote, Karnataka, India

Abstract - The demand for ultra-high frame rates and realistic graphics has driven the evolution of rendering techniques in real-time graphics. NVIDIA's Deep Learning Super Sampling (DLSS) technology addresses performance constraints by leveraging AI-powered models to reconstruct or generate high-resolution frames. The latest iteration, DLSS 4, introduces Multi Frame Generation (MFG) with Transformer- based architectures to synthesize multiple future frames with high temporal and spatial fidelity. This paper explores the evolution, technological framework, implementation, and performance of DLSS 4, comparing it with competing technologies like AMD FSR and Intel XeSS.

Keywords: Frame generation, DLSS, Multi Frame Generation, Transformer, Ray Reconstruction, AI rendering.